Micro-level porosimetry of virtual cementitious materials: structural impact on mechanical and durability evolution

Abstract

Understanding the microstructure of cement paste is the basis of a study towards properties and behaviour of cementi¬tious materials. It is attractive exploit¬ing modern computer facilities for this purpose, favourably competing with time-consuming and laborious experimental approaches. This study aims at bringing mate¬rial studies into virtual reality through a compre¬hensive computa¬tional frame-work that is composed of three parts as de¬scribed below. The first part deals with generating virtual representations of hardening cement composites at micro-scale, starting with producing a paste of randomly packed cement grains at the fresh state. A DEM-based dynamic packing process is used for this purpose to obtain, not only paste with high density but also that with a wide particle size range. The next stage involves simulation of the microstructure during hydration, based on an improved version of the well-known vector approach. The proposed model denoted ‘eXtended Integrated Particle Kinetics Method’ (XIPKM) includes the following improvements: a multi-com¬ponent particle model to take major cement compounds and the pozzolan into account, a numerical technique to capture the complex contact between expanding particles (a crucial issue in vector approaches), and finally a concept to avoid the extreme computational effort in generating a very large amount of fine particles. Furthermore, a numerical procedure is proposed to obtain the basic penetration rates of different minerals instead of using a laborious calibration process commonly used in vector approaches. In the second part, two computational porosimetry methods to explore the pore network characteristics are developed. The first method denoted ‘Random Node Structuring’ (RaNoS) charac¬terises the pore space, based on analysing the configuration of a system of random points dispersed in the pore space. These random points are further employed, together with an enhanced version (for a more efficient size assessment of irregular pores) of the well-known stereological technique – star volume measure (SVM), to estimate the pore size distribution. The second porosimetry method named ‘Double-Random Multiple Tree Structuring’ (DRaMuTS) is an enhanced version of RaNoS, whereby the topology of the pore structure is further efficiently explored by a system of concurrent virtual trees growing and branching randomly in pore space, configured by a robotics-inspired path planning algorithm. Based on topological information attained by the tree systems, the pore space is then converted into a cylindrical tube network for directly estimating permeability. Based on the pore configuration obtained by the porosimetry methods, 2D representative samples to study the tensile damage response of porous materials in bulk as well as interfacial transition zone (ITZ) are proposed, whereby extremely demanding 3D FEM modelling is dis¬missed but the impact of the 3D pore space is nevertheless taken into account. In the final part, several tests are carried out on cement pastes with/without blended pozzolanic admixtures, i.e. rice husk ash (RHA) by applying the presented methodologies, aiming at assessment of the impact of different design parameters (e.g., w/c, cement fineness and RHA blending percentage) on pore characteristics, permeability and tensile damage behaviour. The relations between the pore structural features are discussed. Furthermore, the minimum size for exist¬ence of a representative volume element (RVE) for various pore characteristics as well as tensile damage response is also studied. RHA-blending is shown to improve transport-based capacities but causes a reduction in Young’s modulus, in tensile strength and ductility.